real-ESRGAN

Real-ESRGAN是一种超分辨率网络，全称为Enhanced Super-Resolution Generative Adversarial Networks。该网络通过训练以PSNR为导向的Real-ESRNet模型，然后以Real-ESRNet的网络参数进行初始化，同时使用L1 loss、perceptual loss和GAN loss作为损失函数来训练最终的Real-ESRGAN网络。

Real-ESRGAN被用作一种图像分辨率修复工具，可以提升照片和动画图片的分辨率。它内置了一个预训练模型，能够将分辨率提高到原来的4倍。虽然Real-ESRGAN是一个命令行工具，但使用起来非常简单，而且效果也非常好。<span class="em">1</span><span class="em">2</span><span class="em">3</span>
#### 引用[.reference_title]
- *1* *2* [Real-ESRGAN超分辨网络](https://blog.csdn.net/weixin_46515047/article/details/119829454)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v93^chatsearchT3_2"}}] [.reference_item style="max-width: 50%"]
- *3* [Real-ESRGAN 深度学习超分辨率重建工具](https://download.csdn.net/download/hq10086/79831404)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v93^chatsearchT3_2"}}] [.reference_item style="max-width: 50%"]
[ .reference_list ]